Periodic Reporting for period 1 - ELVIS (ELEM Virtual Heart Populations for Supercomputers)
Período documentado: 2023-03-01 hasta 2024-02-29
In-silico trials are emerging as powerful tools, simulating therapy effects on virtual populations before real-world testing. Providing diverse, expansive populations far exceeding the capacity of traditional clinical trials they have potential to account for broader demographics and genetic variations, leading to more generalizable results. Additionally virtual populations allow for precise control over variables, unlike in vivo trials, and mitigate the ethical concerns surrounding animal and human trials. The ELVIS platform integrates different components to make this a reality at the hand of the biomedical stakeholders.
ELEM’s main product, V.Heart is a supercomputer-based platform to perform massive in-silico clinical trials on populations of digital avatars generated through a database of real medical data, to study the outcomes of different therapies. It assesses the safety and efficacy of novel therapeutics, devices or drugs, by computing thousands of scenarios. Current capability covers classic or leadless pacing and cardiac pumps and cardiac safety of drugs. Our plan is to add new therapies whilst increasing the range of our patented virtual population technology. V.Heart generates unparalleled medical insights and new evidence for biomedical professionals whilst shortening time to market and reducing business and patient risks. It narrows the scope for animal testing and tapers real human trials by evaluating clinical study endpoints much faster and earlier in the development process. Eventually, digital avatars become a Digital Twin of any given patient used as a predictive tool for precision medicine.
ELEM’s main product, V.Heart is a supercomputer-based platform to perform massive in-silico clinical trials on populations of digital avatars generated through a database of real medical data, to study the outcomes of different therapies. It assesses the safety and efficacy of novel therapeutics, devices or drugs, by computing thousands of scenarios. Current capability covers classic or leadless pacing and cardiac pumps and cardiac safety of drugs. Our plan is to add new therapies whilst increasing the range of our patented virtual population technology. V.Heart generates unparalleled medical insights and new evidence for biomedical professionals whilst shortening time to market and reducing business and patient risks. It narrows the scope for animal testing and tapers real human trials by evaluating clinical study endpoints much faster and earlier in the development process. Eventually, digital avatars become a Digital Twin of any given patient used as a predictive tool for precision medicine.
ELEM´s V. Heart platform integrates different software components. The main components of the ELVIS platform are: the WebApp (frontend and backend) which integrates the following two, the Virtual Population Generator and the Simulation Engine based on Alya Red.
We can report main achievements for each of the components’ development.
The Elvis WebApp
After conducting a comprehensive security and risk assessment with one of our customers, we successfully deployed the first production version of V.Heart our cloud-based product designed for cardiotoxicity studies. The deployment is done in commercial clouds: the WebApp is in Amazon Web Services, with the HPC compute nodes in Oracle Cloud Infrastructure. Access security is enforced through a two-step authentication process for customer access. The customer is now using V.Heart in production to study three of its proprietary compounds. Additionally, we are now moving all the results of our cardiotoxicity studies to V.Heart to carry on the analysis directly in the platform.
The simulation engine, Alya Red
At this moment of the project, Alya Red is fully operational to model in-sílico clinical trials cases of cardio toxicity. The simulation engine can efficiently solve large 3D models, each of them representing a virtual patient of a synthetically generated cohort. In the present version of the platform we can simulate the cardiac electrophysiology of healthy patients, together with a new implementation and model parameterization of diseased patients to better represent real populations. So far, the implemented hypertrophic cardiomyopathy in Alya includes
- Implemented ischemia and ischemic material in Alya
- Implemented heart failure in Alya
- Implemented dilated cardiomyopathy in Alya
Additionally, we have expanded the models in two directions: the electromechanical coupling, adding new circulatory system models and improving the coupling itself; and the atrial model, towards a 4-chamber cardiac model.
The Virtual Population Generator
The following was developed:
- Statistical Shape Model, currently applied, but not limited, to biventricular cardiac anatomies. This is a statistical model that allows generating random shapes similar to the ones observed in the dataset used to build the model. In this particular case the model allows us to generate random biventricular cardiac anatomies for the populations. The methodology is based on Cootes et al 1995.
- Improved rule based fiber generation algorithm that includes geometry independent cardiac coordinates to represent any patient information in a standardised way as to be able to compare that information between different patients or transfer it from one patient to another. In particular this methodology allows us to define a template cardiac anatomy whereupon all the patient information can be stored and using which statistical models can be built and statistical analysis can be carried out.
- An automatic pipeline to prepare the computational mesh for the in silico trial. The pipeline generates a high quality computational tetrahedral mesh and adds all the necessary information to the computational mesh such as myocardial fiber orientation, myocardial cell heterogeneity, activation locations simulating the natural activation of the human heart, and the location of the ECG leads for pseudo-ECG measurement.
A way of defining a population starting from a given heart, by changing the myocyte phenotype, material properties and activation patterns. This allows us to generate a wide range of hearts (sharing the same computational mesh) including a variety of cardiac diseases.
This year much focus has been made on identifying and implementing the steps necessary to produce the populations. For the next year we plan to pass the pipeline to one of the publicly available process orchestration solutions, in particular Apache Airflow to allow for better error control and improved efficiency of managing the processes. This will be complemented by adding comprehensive testing and integration into Elem’s gitlab CI/CD workflows.
We can report main achievements for each of the components’ development.
The Elvis WebApp
After conducting a comprehensive security and risk assessment with one of our customers, we successfully deployed the first production version of V.Heart our cloud-based product designed for cardiotoxicity studies. The deployment is done in commercial clouds: the WebApp is in Amazon Web Services, with the HPC compute nodes in Oracle Cloud Infrastructure. Access security is enforced through a two-step authentication process for customer access. The customer is now using V.Heart in production to study three of its proprietary compounds. Additionally, we are now moving all the results of our cardiotoxicity studies to V.Heart to carry on the analysis directly in the platform.
The simulation engine, Alya Red
At this moment of the project, Alya Red is fully operational to model in-sílico clinical trials cases of cardio toxicity. The simulation engine can efficiently solve large 3D models, each of them representing a virtual patient of a synthetically generated cohort. In the present version of the platform we can simulate the cardiac electrophysiology of healthy patients, together with a new implementation and model parameterization of diseased patients to better represent real populations. So far, the implemented hypertrophic cardiomyopathy in Alya includes
- Implemented ischemia and ischemic material in Alya
- Implemented heart failure in Alya
- Implemented dilated cardiomyopathy in Alya
Additionally, we have expanded the models in two directions: the electromechanical coupling, adding new circulatory system models and improving the coupling itself; and the atrial model, towards a 4-chamber cardiac model.
The Virtual Population Generator
The following was developed:
- Statistical Shape Model, currently applied, but not limited, to biventricular cardiac anatomies. This is a statistical model that allows generating random shapes similar to the ones observed in the dataset used to build the model. In this particular case the model allows us to generate random biventricular cardiac anatomies for the populations. The methodology is based on Cootes et al 1995.
- Improved rule based fiber generation algorithm that includes geometry independent cardiac coordinates to represent any patient information in a standardised way as to be able to compare that information between different patients or transfer it from one patient to another. In particular this methodology allows us to define a template cardiac anatomy whereupon all the patient information can be stored and using which statistical models can be built and statistical analysis can be carried out.
- An automatic pipeline to prepare the computational mesh for the in silico trial. The pipeline generates a high quality computational tetrahedral mesh and adds all the necessary information to the computational mesh such as myocardial fiber orientation, myocardial cell heterogeneity, activation locations simulating the natural activation of the human heart, and the location of the ECG leads for pseudo-ECG measurement.
A way of defining a population starting from a given heart, by changing the myocyte phenotype, material properties and activation patterns. This allows us to generate a wide range of hearts (sharing the same computational mesh) including a variety of cardiac diseases.
This year much focus has been made on identifying and implementing the steps necessary to produce the populations. For the next year we plan to pass the pipeline to one of the publicly available process orchestration solutions, in particular Apache Airflow to allow for better error control and improved efficiency of managing the processes. This will be complemented by adding comprehensive testing and integration into Elem’s gitlab CI/CD workflows.
In the context of the Engine Simulation (Alya Red) the main objectives has been the modelling features. The cardiac cycle model has been improved, creating a new, more advanced version of the circulatory system. With this, the Platform will be able to work with a wider range of patient conditions. Also, it has been implemented a model of computational electrophysiology to test new diseases and an atrial electrophysiology model have been integrated, on the way to have a 4-chambers cardiac model.
Regarding the Virtual Population generator, the focus has been on testing arrhythmogenic risk of drugs. The main results in this area made it possible to automate tasks from the previously defined pipeline (mesh morphing algorithm, myocardial fiber generation algorithm and development of the universal biventricular coordinates)
Regarding the Virtual Population generator, the focus has been on testing arrhythmogenic risk of drugs. The main results in this area made it possible to automate tasks from the previously defined pipeline (mesh morphing algorithm, myocardial fiber generation algorithm and development of the universal biventricular coordinates)